Memristor-integrated passive crossbar arrays (CAs) could potentially accelerate neural network (NN) computations, but studies on these devices are limited to software-based simulations owing to their poor reliability. Herein, we propose a self-rectifying memristor-based 1 kb CA as a hardware accelerator for NN computations. We conducted fully hardware-based single-layer NN classification tasks involving the Modified National Institute of Standards and Technology database using the developed passive CA, and achieved 100% classification accuracy for 1500 test sets. We also investigated the influences of the defect-tolerance capability of the CA, impact of the conductance range of the integrated memristors, and presence or absence of selection functionality in the integrated memristors on the image classification tasks. We offer valuable insights into the behavior and performance of CA devices under various conditions and provide evidence of the practicality of memristor-integrated passive CAs as hardware accelerators for NN applications.

Electrical endurance characteristic of resistive switching MnO_x thin film was investigated associated with various oxygen concentrations. From experimental results of various top electrode changing on the examined devices and oxygen concentration during the post-deposition annealing process, it was revealed that electrical endurance characteristic can be significantly improved by possessing high “non-lattice oxygen” concentration in resistive switching thin film and minimizing out-diffusion of oxygen during resistive switching. Finally, a 250 nm-diameter via-hole structure device, composed of TiN/MnO_x/Pt, was fabricated and the promising electrical endurance and retention characteristics and the impressively narrow distribution of resistive switching operation parameters were confirmed in the MnO_x thin film.

Compared with conventional silicon-based semiconductors, amorphous oxide semiconductors present several advantages, including the possibility of room-temperature fabrication, excellent uniformity, high transmittance, and high electron mobility. Notably, the application of oxide semiconductors to flexible electronic devices requires a low-temperature fabrication process. However, for the realization of semiconductor characteristics and stable products, the fabrication process requires annealing at temperatures of 300 °C or higher. To address this, a low-temperature microwave annealing method, which improves the electrical characteristics of a transistor and reduces the production time compared with the conventional annealing method, is presented herein. Microwave annealing is a well-known method of annealing that minimizes the heat energy transferred to a substrate via instantaneous heat transfer through the vibrations of the lattice in the material during microwave irradiation and is suitable as a low-temperature annealing method. In this study, we evaluate the electrical characteristics of devices subjected to conventional annealing at 200 °C and 300 °C for 1 h and microwave annealing at 200 °C for 10 min. For the device subjected to microwave annealing at 200 °C for 10 min, the threshold voltage, current on/off ratio, subthreshold swing, and saturation mobility are 13.9 V, 1.14 × 10⁵, 3.05 V/dec, and 4.23 cm²/V$·$s, respectively. These characteristic results are far superior to the characteristic results of the device subjected to conventional annealing at 200 °C for 1 h and are equivalent to those of the device treated at 300 °C for 1 h. Thus, this study develops a more effective annealing method, which facilitates low-temperature fabrication in a reduced period.

In this research, 1-diode (1D), 1-resistor (1R), and 1D1R stacked devices were separately fabricated using a 400 nm diameter hole substrate. It was observed that, in 1D1R, there was a fivefold increase in endurance and 52% improvement in the resistance distribution characteristics compared to those of 1R. It could be surmised that the stacked diode not only plays the role of a selection device to minimize the interference in the crossbar array type resistive switching device but also acts as an external load resistor to suppress the unexpected current overflow during resistance transition. These experimental results demonstrated that the optimized diode shows significant promise as a unidirectional resistive switching device for memory-related applications.